基于ANOVA和BO-SVM的变压器故障诊断方法

Fault Diagnosis Method of Transformer Based on ANOVA and BO-SVM

为了提高变压器故障诊断模型的诊断精度,提出了一种基于方差分析(analysis of variance, ANOVA)和黑猩猩算法优化支持向量机(bonobo optimizer-support vector machine, BO-SVM)的变压器故障诊断方法。首先,采用方差分析对模型输入量,即变压器油中溶解气体特征进行筛选降维。其次,用黑猩猩算法对影响支持向量机诊断性能的2个参数(核参数与惩罚因子)进行寻优。最后,利用该文所提方法对变压器进行故障诊断,实例仿真结果表明:与IEC、Rogers法相比,采用ANOVA对模型输入量进行筛选降维能更好地提升模型性能;BO算法优化后的支持向量机对比网格搜索算法优化后的支持向量机,在训练速度与诊断精度两方面都有所提升;与粒子群优化支持向量机(particle swarm optimization-support vector machine,PSO-SVM)和遗传算法优化支持向量机(genetic algorithm-support vector machine, GA-SVM)相比,所提方法收敛更快,且故障诊断精度更高、更稳定,BO-SVM、PSO-SVM和GA-SVM的平均诊断正确率分别为91.69%、83.29%、81.34%,验证了所提方法的优越性。

In order to improve the diagnostic accuracy of transformer fault diagnosis model,this paper proposes a transformer fault diagnosis method based on analysis of variance(ANOVA)and bonobo optimizer-support vector machine(BO-SVM).Firstly,ANOVA is used to filter and reduce the dimensionality of the model input quantity(namely,the dissolved gas characteristics)in transformer oil.Secondly,the bonobo optimizer is used to optimize the two parameters(kernel parameter and penalty factor)that affect the diagnostic performance of support vector machine.Finally,the proposed method is used for transformer fault diagnosis.The simulation results show that,compared with IEC and Rogers methods,using ANOVA for screening and dimensionality reduction of the model input can improve model performance.Compared with support vector machine optimized by grid search algorithm,BO-SVM improves both the training speed and the diagnostic accuracy.Compared with particle swarm optimization-support vector machine(PSO-SVM)and genetic algorithm-support vector machine(GA-SVM),the proposed method has faster convergence rate,higher diagnostic accuracy,and more stability.The average diagnostic accuracy of BO-SVM,PSO-SVM and GA-SVM are 91.69%,83.29%and 81.34%,respectively,and the superiority of the proposed method is verified.